Wind turbine blade crack detection method and system using acoustic emission signals

Abstract

The application provides a wind turbine blade crack detection method and system using acoustic emission signals, relates to the technical field of nondestructive testing, and obtains acoustic emission waveform sequences of a wind turbine blade collected by multiple sensors in a preset time window through acquisition of acoustic emission signals of the wind turbine blade collected by the multiple sensors in the preset time window; time delay estimation is performed on the acoustic emission waveform sequences by using cross-correlation to obtain an attenuation coefficient matrix; coordinate calculation is performed on a time difference of arrival sequence by using hyperbolic positioning to obtain an energy distribution matrix; waveform decomposition is performed on acoustic emission waveform sequences corresponding to abnormal paths in the energy distribution matrix by using wavelet transform to obtain a time delay sequence; the coordinate position of a scattering wave signal is determined according to the time delay sequence, and the physical boundary of a layered crack in the wind turbine blade is determined. The high-precision accurate positioning and three-dimensional form restoration of the physical boundary of the hidden layered crack in the wind turbine blade are realized.

Description

Technical Field

[0001] This application relates to the field of nondestructive testing technology, and in particular to a method and system for detecting cracks in wind turbine blades using acoustic emission signals. Background Technology

[0002] With the rapid growth of global wind power installations, the blades, as key components for capturing wind energy in wind turbines, are directly related to the economic benefits and lifespan of the units due to their operational safety. Because blades are subjected to harsh alternating loads for extended periods, they are highly susceptible to structural damage such as delamination and cracks. Therefore, real-time and accurate health monitoring of blades has become a crucial issue in the field of wind power operation and maintenance.

[0003] Existing methods typically deploy sensor networks on the surface of wind turbine blades to perform structural assessments by capturing elastic wave signals excited by material damage. These methods often employ threshold triggering or time-of-flight localization algorithms to calculate the location coordinates of the acoustic emission source and combine this with amplitude characteristics to make a preliminary determination of the defect level.

[0004] However, due to the extremely strong non-uniformity and anisotropy of blade composite materials, the dynamic attenuation law of stress waves during propagation is extremely complex, leading to significant biases in the assessment of the spatial distribution of defect energy using existing methods. Furthermore, traditional monitoring methods primarily focus on point location of the sound source, making it difficult to effectively separate and utilize complex reflected signals to represent the geometry of the damaged area, thus failing to accurately reconstruct the physical contours of the defect. Therefore, existing technologies face the technical challenge of accurately and quantitatively characterizing the extent of damage within the blade. Summary of the Invention

[0005] The purpose of this application is to provide a method and system for detecting cracks in wind turbine blades using acoustic emission signals, in order to solve the technical problem in the prior art that it is difficult to accurately and quantitatively characterize the extent of damage inside the blade.

[0006] In a first aspect, this application provides a method for detecting cracks in wind turbine blades using acoustic emission signals, including: Acoustic emission signals of wind turbine blades collected by multiple sensors within a preset time window are acquired, and each acoustic emission signal is synchronously converted from analog to digital to obtain an acoustic emission waveform sequence. Cross-correlation is used to estimate the time delay of the acoustic emission waveform sequence to obtain the arrival time difference sequence. Then, distance weighting and gain compensation are performed on multiple peak amplitudes extracted from the acoustic emission waveform sequence to obtain the attenuation coefficient matrix. Hyperbolic positioning is used to calculate the coordinates of the time difference of arrival sequence to obtain the sound source coordinates, and the propagation distance sequence between the sound source coordinates and the sensor is calculated. By curve fitting the propagation distance sequence and the attenuation coefficient matrix, the energy distribution matrix is ​​obtained. Wavelet transform is used to decompose the acoustic emission waveform sequence corresponding to the abnormal path in the energy distribution matrix to obtain the main wave signal and the scattered wave signal. The time offset of the scattered wave signal relative to the main wave signal is calculated to obtain the time delay sequence. The abnormal path is determined based on the energy abrupt change point in the energy distribution matrix. The coordinates of the scattered wave signal are determined based on the time delay sequence. The energy mutation point and the coordinates are correlated and projected onto the pre-constructed blade model for boundary fitting to determine the physical boundary of the layered crack in the wind turbine blade.

[0007] Optionally, after determining the coordinate position of the scattered wave signal based on the time delay sequence, and projecting the energy abrupt change point and coordinate position onto a pre-built blade model for boundary fitting to determine the physical boundary of the delamination crack in the wind turbine blade, the method further includes: The target peak amplitude of all scattered wave signals corresponding to the abnormal path is counted, and each target peak amplitude is divided into a preset amplitude interval. The number of scattered waves in each amplitude interval is counted to obtain the frequency sequence. The slope of the damage distribution is obtained by linearly fitting the number of scattered waves and the corresponding peak amplitude in the frequency sequence. The slope of the damage distribution is compared with the preset crack classification threshold to determine the crack damage type corresponding to the delamination crack, and the crack damage type is marked to the physical boundary. The crack damage types include matrix cracking, fiber debonding and fiber breakage.

[0008] Optionally, the time delay of the acoustic emission waveform sequence is estimated using cross-correlation to obtain the arrival time difference sequence, including: The channel with the earliest arrival time in the acoustic emission waveform sequence is used as the reference channel, the waveform corresponding to the reference channel is used as the reference waveform, and the remaining channels in the acoustic emission waveform sequence are used as the channels to be tested, and the channels corresponding to the channels to be tested are used as the waveforms to be tested. Each waveform to be measured is translated, and the degree of overlap between the envelope of each waveform to be measured and the reference waveform is compared to obtain the relative displacement of each waveform to be measured. Each relative displacement is multiplied by a preset sampling period to obtain the relative time delay value of each waveform under test. All relative time delay values ​​are then arranged according to the channel order in the acoustic emission waveform sequence to obtain the arrival time difference sequence.

[0009] Optionally, distance weighting and gain compensation are performed on multiple peak amplitudes extracted from the acoustic emission waveform sequence to obtain an attenuation coefficient matrix, including: The highest voltage point of each channel in the acoustic emission waveform sequence is taken as the peak amplitude of the waveform of each channel; The relative time delay value in the arrival time difference sequence is multiplied by the preset wave velocity value to obtain the distance deviation value of each channel under test relative to the reference channel. Based on each distance deviation value and the spatial position of the corresponding sensor, multiple sets of positioning trajectory lines are determined, and the intersection points of multiple sets of positioning trajectory lines are determined as the initial sound source of the acoustic emission signal. Calculate the geometric distance between the initial sound source and each sensor, and based on the geometric distance, correct each peak amplitude to obtain the corrected peak amplitude; The quotient between the corrected peak amplitude of the reference channel and the corrected peak amplitude of each channel under test is calculated to obtain the attenuation coefficient matrix.

[0010] Optionally, hyperbolic positioning is used to calculate the coordinates of the time difference of arrival sequence to obtain the sound source coordinates, and the propagation distance sequence between the sound source coordinates and the sensor is calculated. By curve fitting the propagation distance sequence and the attenuation coefficient matrix, the energy distribution matrix is ​​obtained, including: Based on the time difference of arrival sequence, multiple sets of hyperbolic trajectory equations are constructed with any two sensors as the focus, and multiple candidate points are selected within the area determined by the multiple sets of hyperbolic trajectory equations with the initial sound source as the search center. Calculate the theoretical time delay difference between each candidate point and the sensor corresponding to the channel under test and the sensor corresponding to the reference channel. Calculate the difference between each theoretical time delay difference and the relative time delay value of the corresponding channel in the arrival time difference sequence to obtain the residual corresponding to each candidate point. Then, take one or more candidate points corresponding to the minimum residual as the intersection point. Calculate the spatial centroid of all intersection points to obtain the sound source coordinates, calculate the straight line length between the sound source coordinates and each sensor to obtain the propagation distance sequence, and construct a fitting function with each straight line length in the propagation distance sequence as the independent variable and the quotient of the corresponding channel in the attenuation coefficient matrix as the dependent variable. Using the spatial location of each grid node in the pre-built blade model as a sampling point, the coordinates corresponding to each sampling point are substituted into the fitting function for assignment and mapping to obtain the energy distribution matrix.

[0011] Optionally, wavelet transform is used to decompose the acoustic emission waveform sequence corresponding to the abnormal path in the energy distribution matrix to obtain the main wave signal and the scattered wave signal. The time offset of the scattered wave signal relative to the main wave signal is then calculated to obtain the time delay sequence, including: The numerical differences between adjacent coordinate points in the energy distribution matrix are calculated to obtain multiple numerical change rates. The coordinate points corresponding to the numerical change rates that exceed the preset deviation are identified as energy mutation points. Based on the structural discontinuity of the region where each energy mutation point is located, the reflection interface corresponding to each energy mutation point is determined. The line connecting the sound source coordinates to each sensor is used as a candidate path, and the candidate path that intersects with any energy mutation point is determined as an abnormal path. Frequency band layering is performed on the acoustic emission waveform sequence corresponding to each abnormal path to obtain multiple wave packet components corresponding to each abnormal path. The earliest arriving wave packet component in each abnormal path is taken as the main wave signal, and the wave packet component corresponding to the reflection interface is taken as the scattered wave signal. The starting trigger time of the scattered wave signal and the main wave signal under each abnormal path are compared to obtain the time offset corresponding to each abnormal path. Then, all time offsets are arranged into a time delay sequence according to the channel order in the acoustic emission waveform sequence.

[0012] Optionally, the coordinate positions of the scattered wave signals are determined based on the time delay sequence, and the energy abrupt change points and coordinate positions are correlated and projected onto a pre-built blade model for boundary fitting to determine the physical boundaries of the delamination cracks in the wind turbine blade, including: Calculate the product of each time offset in the time delay sequence and the preset sound speed to obtain the travel increment corresponding to each scattered wave signal; Based on each stroke increment, multiple elliptical trajectories are constructed with the sound source coordinates and the corresponding sensor position as the focus, and the intersection of multiple elliptical trajectories is used as the coordinate position of each scattered wave signal. Establish the correlation between each energy mutation point and each coordinate position, and match each coordinate point that meets the correlation to the nearest grid node in the blade model to obtain the projected coordinates of each grid node. Determine the target connected region based on the projected coordinates. The outermost edge of the target connected region is contour traced to obtain a closed line, and the geometric region enclosed by the closed line is used as the physical boundary.

[0013] Secondly, this application provides a wind turbine blade crack detection system utilizing acoustic emission signals, comprising: The acquisition module is used to acquire acoustic emission signals of wind turbine blades collected by multiple sensors within a preset time window, and to perform synchronous analog-to-digital conversion on each acoustic emission signal to obtain an acoustic emission waveform sequence. The estimation module is used to estimate the time delay of the acoustic emission waveform sequence using cross-correlation to obtain the arrival time difference sequence, and to calculate the distance weight and gain compensation of multiple peak amplitudes extracted from the acoustic emission waveform sequence to obtain the attenuation coefficient matrix. The calculation module is used to calculate the coordinates of the time difference of arrival sequence using hyperbolic positioning to obtain the sound source coordinates, and calculate the propagation distance sequence between the sound source coordinates and the sensor. By curve fitting the propagation distance sequence and the attenuation coefficient matrix, the energy distribution matrix is ​​obtained. The decomposition module is used to decompose the acoustic emission waveform sequence corresponding to the abnormal path in the energy distribution matrix using wavelet transform to obtain the main wave signal and the scattered wave signal, and calculate the time offset of the scattered wave signal relative to the main wave signal to obtain the time delay sequence. The abnormal path is determined based on the energy abrupt change point in the energy distribution matrix. The fitting module is used to determine the coordinate position of the scattered wave signal based on the time delay sequence, and then project the energy mutation point and coordinate position onto the pre-built blade model for boundary fitting to determine the physical boundary of the delamination crack in the wind turbine blade.

[0014] Thirdly, this application provides an electronic device, comprising: Memory, used to store computer programs; A processor is used to execute a computer program to implement the steps of the wind turbine blade crack detection method using acoustic emission signals as described in the first aspect above.

[0015] Fourthly, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, can implement the steps of the wind turbine blade crack detection method using acoustic emission signals as described in the first aspect above.

[0016] The wind turbine blade crack detection method utilizing acoustic emission signals provided in this application ensures high consistency of the original acoustic signal over time through multi-channel synchronous acquisition and analog-to-digital conversion within a preset time window. This resists background noise interference in the blade operating environment and restores the true physical parameters representing the degree of damage. It achieves precise point-to-point localization of the damage initiation location, mitigating the bias in energy spatial distribution assessment in existing technologies. Key time-dependent information carrying the geometric features of the damage is extracted from complex reflection signals. This represents a leap from point localization to surface characterization, achieving precise quantitative reconstruction of the physical contour of the damaged area.

[0017] Furthermore, this application determines the rate of change of values ​​by calculating the numerical differences between adjacent coordinate points in the energy distribution matrix, thereby identifying the energy abrupt change points characterizing structural discontinuities and their corresponding reflection interfaces. Among the candidate paths formed by connecting the sound source coordinates to each sensor, paths intersecting with the energy abrupt change points are identified as anomalous paths. Subsequently, the signal sequences corresponding to the anomalous paths are band-separated, and the main wave signal and scattered wave signal are separated based on arrival time and reflection interface characteristics. Finally, by comparing the initial trigger times of the two and arranging them in channel order, a time delay sequence reflecting damage depth and path characteristics is formed. This solves the problems of blindness and unreliability in traditional waveform decomposition when processing reflection signals from composite blades. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 A schematic flowchart illustrating the wind turbine blade crack detection method utilizing acoustic emission signals provided in this application embodiment; Figure 2 A flowchart illustrating the method for obtaining a time delay sequence provided in an embodiment of this application; Figure 3 A flowchart illustrating the method for determining the physical boundary of delamination cracks in wind turbine blades provided in this application embodiment; Figure 4 A schematic diagram of the structure of a wind turbine blade crack detection system utilizing acoustic emission signals provided in an embodiment of this application; Figure 5 This is a schematic diagram of the hardware structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0020] To address the extreme complexity of stress wave propagation dynamic attenuation caused by the strong non-uniformity of composite materials in wind turbine blades, and the limitations of existing technologies that can only locate sound source points but cannot quantitatively characterize damage boundaries, this application constructs an energy distribution matrix that can objectively eliminate material anisotropic interference and reflect the true attenuation law by performing distance weighting and gain compensation on the extracted peak amplitude. This solves the problem of deviation in the spatial assessment of damage energy in existing methods. Furthermore, by using the energy mutation points in this matrix as geometric priors, the technical bottleneck of traditional methods being unable to effectively separate complex reflection signals is broken. The scattered wave signals characterizing the crack interface are accurately extracted through wavelet transform. Finally, by associating and mapping the spatial coordinates of the scattering source with the energy mutation location and projecting them onto the pre-constructed blade model, a dimensional leap from discrete sound source point location to continuous physical boundary characterization is achieved, solving the technical problem of the difficulty in accurately and quantitatively characterizing the range of damage areas inside the blade in existing technologies.

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] The core of this application is to provide a method for detecting cracks in wind turbine blades using acoustic emission signals, and a flowchart of one specific implementation is shown below. Figure 1 As shown, the method includes: Step 101: Acquire acoustic emission signals of wind turbine blades collected by multiple sensors within a preset time window, perform synchronous analog-to-digital conversion on each acoustic emission signal, and obtain an acoustic emission waveform sequence.

[0023] In this step, the preset time window refers to the time interval used to limit the duration of acoustic emission signal acquisition. A sensor is an energy conversion device installed on the surface of a wind turbine blade to sense stress waves; its placement can include the main spar or the tip of the blade, a key monitoring area. A wind turbine blade is the energy-harvesting component of a wind turbine made of composite materials and is also the physical entity for monitoring delamination cracks. An acoustic emission signal refers to the transient elastic wave generated by stress release due to the initiation or propagation of internal cracks in the blade. An acoustic emission waveform sequence is a discrete set of values ​​formed after synchronous sampling and quantization of the acoustic emission signal, used to represent the signal strength characteristics that change over time.

[0024] In this embodiment of the application, firstly, by arranging on the wind turbine blades Multiple sensors at different coordinate points on the surface The system senses elastic waves generated within the structure in real time. When delamination damage occurs inside the blade due to alternating loads, acoustic emission signals are generated. Then, within a preset time window... Inside, each sensor converts the received mechanical vibration into a continuous analog voltage signal. Then, a synchronized analog-to-digital conversion is performed on the analog voltage signal of each channel using a uniform sampling trigger pulse.

[0025] Specifically, by sharing a high-precision sampling clock signal to simultaneously drive multiple parallel analog-to-digital converter chips, it is ensured that each converter chip instantaneously samples and binary-quantizes the analog voltage corresponding to its respective channel at perfectly consistent nanosecond intervals. This results in the generated multi-channel digital information having no phase shift or clock jitter deviation on the time axis. Ultimately, a sequence of acoustic emission waveforms composed of digital sampled values ​​can be obtained. For example, for sensors... The obtained acoustic emission waveform sequence .

[0026] Step 102: Use cross-correlation to estimate the time delay of the acoustic emission waveform sequence to obtain the arrival time difference sequence, and perform distance weight calculation and gain compensation on multiple peak amplitudes extracted from the acoustic emission waveform sequence to obtain the attenuation coefficient matrix.

[0027] In this step, cross-correlation refers to a mathematical algorithm that assesses similarity and determines time delay by translating two signal sequences and calculating the integral of their product. The time difference of arrival sequence is a vector set consisting of the arrival time delay values ​​of waveforms acquired by multiple sensors relative to a reference waveform. Peak amplitude refers to the maximum voltage value of the acoustic emission waveform sequence within a specific time period. The attenuation coefficient matrix is ​​a set of values ​​reflecting the proportion of energy loss due to geometric diffusion and material absorption along the propagation path of the sound wave.

[0028] Step 201: Take the channel with the earliest arrival time in the acoustic emission waveform sequence as the reference channel, take the waveform corresponding to the reference channel as the reference waveform, take the remaining channels in the acoustic emission waveform sequence as the channels to be tested, and take the channel corresponding to the channel to be tested as the waveform to be tested.

[0029] In this step, the reference channel refers to the sensor channel with the earliest trigger time among the multiple acquired channels. The reference waveform refers to the digitized voltage signal sequence acquired by the reference channel. The channel under test refers to the other sensor channels besides the reference channel. The waveform under test refers to the digitized voltage signal sequence acquired by the channel under test.

[0030] In this embodiment of the application, the wind turbine blades are first extracted. Each sensor The acquired acoustic emission waveform sequence is used. Numerical comparison is used to determine the start time of amplitude abrupt changes in each waveform sequence, thus determining the order of triggering moments. For example, when a crack activity is detected, the sensor... The triggering time is ,sensor The triggering time is .like Less than Then the sensor The corresponding channel is determined as the reference channel, and its corresponding reference waveform is obtained. Next, channels 2, 3, and 4 are designated as the channels to be tested, and the corresponding waveforms to be tested are obtained. .

[0031] Step 202: Translate each waveform to be tested and compare the degree of overlap between the envelope of each waveform to be tested and the reference waveform to obtain the relative displacement of each waveform to be tested.

[0032] In this step, the degree of envelope overlap refers to the similarity of the overlap and matching between the envelope of the waveform under test and the envelope of the reference waveform on the time axis. Relative displacement refers to the number of translation sampling points required for the waveform under test to achieve optimal envelope overlap relative to the reference waveform.

[0033] In this embodiment of the application, each waveform to be measured is first... Compared with reference waveform Hilbert transform processing is performed on each waveform. Specifically, demodulation is performed by calculating the magnitude of the analytic signal to extract the low-frequency envelope curve of each waveform. Then, keeping the envelope curve of the reference waveform stationary, the envelope curve of each waveform under test is progressively translated along the time axis with sampling points as the step size. At each translation step, the cross-correlation coefficient between the envelope of the waveform under test and the envelope of the reference waveform is calculated. Finally, the relative displacement of each waveform under test is determined by finding the translation position where the maximum cross-correlation coefficient is located.

[0034] For example, the waveform to be measured in channel 2. The obtained relative displacement is expressed as an integer number of points. .

[0035] Step 203: Multiply each relative displacement by the preset sampling period to obtain the relative time delay value of each waveform to be measured, and arrange all the relative time delay values ​​according to the channel order in the acoustic emission waveform sequence to obtain the arrival time difference sequence.

[0036] In this step, the preset sampling period refers to the time sampling step constant pre-set when the hardware device performs signal quantization. The relative time delay value refers to the time interval between the sound wave arriving at the sensor under test and arriving at the reference sensor.

[0037] In this embodiment of the application, the sampling period preset by the analog-to-digital converter chip is obtained. The relative time delay of each measured waveform relative to the reference waveform is obtained by multiplying the relative displacement and the sampling period. The specific calculation formula is as follows: ,in, Indicates the first The relative time delay value of each waveform to be measured. This indicates the corresponding relative displacement. This indicates the sampling period. Next, all relative delay values ​​are arranged in order of the physical number of the sensor channels.

[0038] For example, we can obtain row vectors representing the arrival time difference sequence. The first digit, 0, represents the time difference of the reference channel itself.

[0039] Step 211: Take the highest voltage point of each channel in the acoustic emission waveform sequence as the peak amplitude of each channel waveform.

[0040] In this step, the voltage peak point refers to the sampling point where the absolute voltage value is largest within the time range of the acoustic emission waveform sequence. The peak amplitude refers to the absolute voltage value corresponding to the voltage peak point.

[0041] In this embodiment, amplitude retrieval processing is performed on the acoustic emission waveform sequence of each acquired channel. First, the absolute value of each sampled value in the sequence is taken to obtain the corresponding set of absolute voltage values. Then, the set is traversed and the maximum value is searched.

[0042] For example, for waveform sequences Extract the sampling point with the largest absolute voltage value as the peak amplitude. Finally, by combining the extraction results from all channels, the peak amplitude vector representing the signal intensity distribution is obtained. .

[0043] Step 212: Multiply the relative time delay value in the arrival time difference sequence with the preset wave velocity value to obtain the distance deviation value of each channel under test relative to the reference channel. Based on each distance deviation value and the spatial position of the corresponding sensor, determine multiple sets of positioning trajectory lines and determine the intersection point of multiple sets of positioning trajectory lines as the initial sound source of the acoustic emission signal.

[0044] In this step, the preset wave velocity value refers to the theoretical velocity constant of sound waves propagating in a specific composite material blade structure. The distance deviation value refers to the difference in the sound wave propagation path length calculated based on the time difference and sound velocity. The positioning trajectory line refers to the geometric line within the blade space that meets the requirement of constant distance deviation. The initial sound source refers to the sound wave emission position point initially calculated using time delay information.

[0045] In this embodiment of the application, a preset wave velocity value is first obtained. Arrival time difference sequence Each relative delay value in With preset wave speed value Multiply by each other to calculate the distance deviation value. Next, combining the three-dimensional spatial coordinates of each sensor, multiple sets of positioning trajectory lines are constructed with the reference sensor and the sensor under test as the focal points. Specifically, each set of positioning trajectory lines corresponds to a hyperbola equation. By solving the simultaneous equations of multiple hyperbolas, the common intersection point is determined as the initial sound source coordinates. .

[0046] Step 213: Calculate the geometric distance between the initial sound source and each sensor, and based on the geometric distance, correct each peak amplitude to obtain the corrected peak amplitude.

[0047] In this step, geometric distance refers to the straight-line length between the initial sound source point and the sensor mounting point. The corrected peak amplitude refers to the amplitude value after propagation attenuation weighting compensation.

[0048] In this embodiment of the application, the initial sound source coordinates are first determined. Three-dimensional values ​​and various sensors Calculate the geometric distance between the two using fixed three-dimensional coordinates. Next, distance attenuation weights are determined based on the spherical wave propagation attenuation law. Then, these weights are used to adjust the distance attenuation for each peak amplitude. Perform the correction process. The specific calculation formula is shown in formula (1): (1) in, This represents the corrected peak amplitude. Indicates the original peak amplitude. This represents the geometric distance. This process eliminates amplitude differences caused by unequal distances between the sensor and the sound source. For example, it yields the corrected peak amplitude vector. .

[0049] Step 214: Calculate the quotient between the corrected peak amplitude of the reference channel and the corrected peak amplitude of each channel under test to obtain the attenuation coefficient matrix.

[0050] In this step, the attenuation coefficient matrix refers to the set of attenuation ratio values ​​of each channel signal arranged in channel order.

[0051] In this embodiment, the corrected peak amplitude corresponding to the reference channel is first extracted. Next, calculate using division operations. Corrected peak amplitude for each channel under test The ratio between them yields the quotient. Then, all quotients are arranged in order of sensor channel number. This results in an attenuation coefficient matrix representing the overall energy attenuation ratio. :

[0052] Since the ratio of the reference channel to itself is 1, the first element of the matrix is ​​fixed at 1.

[0053] Step 103: Use hyperbolic positioning to calculate the coordinates of the time difference of arrival sequence to obtain the sound source coordinates, and calculate the propagation distance sequence between the sound source coordinates and the sensor. By curve fitting the propagation distance sequence and the attenuation coefficient matrix, the energy distribution matrix is ​​obtained.

[0054] In this step, hyperbolic positioning refers to a positioning method that determines the spatial location of the acoustic emission source based on the intersection of hyperbolas determined by the time difference of arrival between different sensors. The sound source coordinates refer to the three-dimensional numerical values ​​of the starting position of the crack acoustic emission signal in the blade coordinate system. The propagation distance sequence refers to the numerical vector composed of the geometric straight-line distances from the sound source coordinates to the installation positions of each sensor. The energy distribution matrix refers to the energy intensity mapping table formed by mapping the discrete attenuation characteristics to the global blade grid after fitting.

[0055] Step 301: Based on the arrival time difference sequence, construct multiple sets of hyperbolic trajectory equations with any two sensors as the focus, and select multiple candidate points within the area determined by the multiple sets of hyperbolic trajectory equations with the initial sound source as the search center.

[0056] In this step, the hyperbolic trajectory equation refers to the surface equation formed by all points in three-dimensional space whose distance difference to two known sensor positions is equal to a constant. The search center refers to the reference position determined with the initial sound source as the center, used to limit the high-precision positioning search range. Candidate points refer to coordinate sampling points extracted within the search area according to the sampling density for residual calculation.

[0057] In this embodiment of the application, the arrival time difference sequence is first extracted. With every two sensors and The position is the focal point, using the preset wave speed The corresponding relative time delay values ​​are converted into distance differences, and multiple sets of hyperbolic trajectory equations are constructed. Next, the preset search radius is obtained. The preset search radius refers to the pre-defined geometric length limit extending outward from the search center. (Based on the initial sound source...) As a search center, in the presence of Within the enclosed local three-dimensional space, discretization sampling is performed according to a preset spatial sampling density. The preset spatial sampling density refers to the constant time or spatial interval between the selected discrete points within the search area. For example, a candidate point vector set consisting of multiple coordinate points can be obtained. .

[0058] Step 302: Calculate the theoretical time delay difference between each candidate point and the sensor corresponding to the channel under test and the sensor corresponding to the reference channel. Calculate the difference between each theoretical time delay difference and the relative time delay value of the corresponding channel in the arrival time difference sequence to obtain the residual corresponding to each candidate point. Then, take one or more candidate points corresponding to the minimum residual as the intersection point.

[0059] In this step, the theoretical time delay difference refers to the predicted time difference between the arrival of the sound wave from the candidate point location to different sensors, calculated based on the geometric path length. The residual refers to the deviation between the theoretical time delay difference and the measured relative time delay value under the same signal channel. The intersection point is the candidate point with the smallest residual value that is closest to the actual sound emission source location.

[0060] In this embodiment of the application, each candidate point is first calculated. To the sensor under test distance and to the reference sensor distance Using formulas Calculate the theoretical time delay difference for each candidate point. Then, assign each theoretical time delay difference... With arrival time difference sequence The relative delay values ​​of the corresponding channels are interpolated to obtain the residual value for each candidate point. The specific residual calculation process is shown in formula (2): (2) in, Indicates the first The residuals of each candidate point Indicates the total number of sensors. Indicates the first The candidate point corresponds to the first... The theoretical time delay difference of each channel, Indicates the first The relative delay values ​​of each channel are calculated. Then, the residual values ​​of all candidate points are iterated over, and the minimum value is searched. Finally, one or more candidate points that achieve the minimum value are determined as the intersection points. .

[0061] Step 303: Calculate the spatial centroid of all intersection points to obtain the sound source coordinates. Calculate the straight line length between the sound source coordinates and each sensor to obtain the propagation distance sequence. Construct a fitting function with each straight line length in the propagation distance sequence as the independent variable and the quotient of the corresponding channel in the attenuation coefficient matrix as the dependent variable.

[0062] In this step, the spatial centroid refers to the average position coordinates calculated from multiple spatial coordinate points according to the geometric center principle. The fitting function is a mathematical regression equation that represents the regular attenuation of signal amplitude intensity as the propagation distance increases.

[0063] In this embodiment, the intersection points of all residuals that satisfy the preset deviation requirements are first extracted. By analyzing all intersection points The spatial coordinate components are averaged to calculate the spatial centroid, which is then used to determine the final sound source coordinates. Next, calculate separately. With each sensor Geometric straight line length between For example, arranging the channels in order yields a propagation distance sequence. Then, in sequence The length of each line in the matrix is ​​the independent variable, and the attenuation coefficient matrix is ​​used as the reference. The quotient of the corresponding channel is the dependent variable, and nonlinear regression analysis is performed.

[0064] Specifically, firstly, a power function regression model is established. To determine the coefficients, the independent variable... With dependent variable Perform the natural logarithmic transformation separately to convert the nonlinear model into a linear model. Next, the least squares method is used to perform linear regression on the transformed data pairs to find the slope that minimizes the sum of squared errors. With intercept Finally, the fitting coefficients are determined through inverse exponential operations. Specifically, a fitting function in the form of a power function can be constructed as shown in formula (3): (3) in, Indicates distance from the sound source The expected decay ratio at that location, and The coefficients are the fitting coefficients. This function accurately reproduces the energy loss law of sound wave propagation in composite blades.

[0065] Step 304: Using the spatial location of each grid node in the pre-built blade model as a sampling point, substitute the coordinates corresponding to each sampling point into the fitting function for assignment and mapping to obtain the energy distribution matrix.

[0066] In this step, sampling points refer to spatial locations extracted from pre-built blade model grid data for energy intensity assessment.

[0067] In this embodiment, a pre-built blade model is first invoked, and discrete mesh nodes inside and on the surface of the model are extracted. Each mesh node is used as a sampling point. Next, the coordinates from each sampling point to the sound source are calculated. straight-line distance After that, Substitute into the fitting function The calculation is performed to obtain the expected energy value corresponding to the sampling point. Then, the deviation between the measured amplitude of each sensor within the sampling point area and the fitted expected value is obtained. Specifically, the corrected amplitude of each discrete sensor is mapped to a global grid using an interpolation algorithm, and then superimposed with the smoothed background trend generated by the fitting function to generate an energy distribution matrix. .matrix The values ​​in the graph reflect the deviation of the local measured energy from the global fitting trend, i.e., the residual energy map.

[0068] For example, in one Within the mapping region, the matrix The form is expressed as:

[0069] in, This represents the energy assignment intensity at the corresponding grid sampling point.

[0070] Step 104: Use wavelet transform to decompose the acoustic emission waveform sequence corresponding to the abnormal path in the energy distribution matrix to obtain the main wave signal and the scattered wave signal, and calculate the time offset of the scattered wave signal relative to the main wave signal to obtain the time delay sequence. The abnormal path is determined based on the energy abrupt change point in the energy distribution matrix.

[0071] In this step, wavelet transform refers to a transformation method that uses basis functions with localization properties to perform multi-scale decomposition of non-stationary signals to extract time-frequency features. Anomaly path refers to a spatial straight line connecting the sound source coordinates and the sensor position, passing through an energy abrupt change point. The dominant wave signal refers to the first wave packet component that propagates directly from the sound source to the sensor. Scattered wave signal refers to the wave packet component whose arrival is delayed due to reflection or diffraction at layered interfaces during propagation. Time offset refers to the time difference between the arrival of the scattered wave signal and the dominant wave signal within the same channel. Time delay sequence refers to the set of multiple time offset values ​​arranged sequentially by channel. An energy abrupt change point refers to a singular location in the energy distribution matrix where the rate of change of the value exceeds a preset deviation, indicating a change in structural characteristics.

[0072] like Figure 2 As shown, Figure 2 This is a flowchart illustrating the method for obtaining a time delay sequence provided in an embodiment of this application.

[0073] Step 401: Calculate the numerical difference between adjacent coordinate points in the energy distribution matrix to obtain multiple numerical change rates, and determine the coordinate points corresponding to the numerical change rates that exceed the preset deviation as energy mutation points. Based on the structural discontinuity of the region where each energy mutation point is located, determine the reflection interface corresponding to each energy mutation point.

[0074] In this step, the numerical rate of change refers to the gradient value of energy intensity between spatially adjacent grid nodes in the energy distribution matrix, used to represent the degree of abrupt change in energy decay. The preset deviation refers to a pre-set numerical threshold used to determine whether a nonlinear abrupt change has occurred in the energy attribute. Structural discontinuity refers to physical irregularities such as delamination or fracture of composite materials within the wind turbine blade. The reflecting interface refers to the geometrical plane formed by structural discontinuities that can cause sound wave reflection.

[0075] In this embodiment, the energy distribution matrix is ​​first extracted. The numerical differential operator is used to calculate the numerical difference between adjacent coordinate points in the matrix. Specifically, this is done by calculating the difference between each grid node and the corresponding coordinate point. direction and The partial derivative in direction yields the rate of change of the value. For example, the calculated numerical rate of change matrix Represented as:

[0076] Next, the matrix Each element in the preset deviation Perform a comparison. If the value exceeds... Then, this coordinate point is determined as the energy mutation point. Next, local space fitting is performed on the identified multiple energy mutation points. Specifically, for each energy mutation point... The algorithm searches for a set of neighboring points within a preset distance range. It then extracts the three-dimensional spatial coordinates of all points in this set and constructs a coordinate observation matrix. By solving for the eigenvectors of this matrix or minimizing the sum of squared distances from all points to the plane using the least squares method, it calculates the plane equation parameters for the local region where the point is located. Finally, these equation parameters are used to determine the reflecting interface representing the physical morphology of the layered defect. .

[0077] Step 402: The line connecting the sound source coordinates to each sensor is taken as a candidate path, and the candidate path that intersects with any energy mutation point is identified as an abnormal path.

[0078] In this step, a candidate path refers to the geometric line connecting the sound wave propagating in a straight line from the sound source coordinates to each sensor installation location under the assumption of an ideal homogeneous medium. An abnormal path refers to a candidate path in actual propagation where the sound wave energy is abnormally attenuated or reflects due to encountering abrupt energy changes.

[0079] In this embodiment of the application, the coordinates of the sound source are first obtained. and the spatial coordinates of each sensor Next, connect them using the equation of a straight line. For each sensor location, multiple candidate paths are obtained. For example, the resulting candidate path vector... Then, spatial geometric collision detection is performed to determine each energy mutation point. Is it located in the candidate path? Within the coverage area. Specifically, if the candidate path With any energy mutation point The existence of an intersection point indicates that beam propagation along that path is obstructed, thus classifying the path as an abnormal path. The final set of anomaly paths pointed to key sensor channels that might capture the scattered signals.

[0080] Step 403: Perform frequency band layering on the acoustic emission waveform sequence corresponding to each abnormal path to obtain multiple wave packet components corresponding to each abnormal path. Take the earliest arriving wave packet component in each abnormal path as the main wave signal and the wave packet component corresponding to the reflection interface as the scattered wave signal.

[0081] In this step, the wave packet component refers to a local signal cluster with a distinct energy envelope characteristic within a specific frequency sub-band. The main wave signal refers to the original beam component that arrives directly at the sensor without being reflected by the interface. The scattered wave signal refers to the secondary beam component with time-delay characteristics after being reflected by the reflecting interface.

[0082] In this embodiment of the application, each abnormal path is first extracted. The corresponding acoustic emission waveform sequence is obtained. A multi-level decomposition operation is performed on the sequence using discrete wavelet transform to complete frequency band layering. Then, multiple wave packet components with energy concentration characteristics are extracted from the decomposed coefficients. For example, a wave packet component sequence obtained for a specific anomalous path is shown. Then, the sequence is arranged according to the order of arrival time. Perform feature recognition. Select the earliest arriving and strongest wave packet component as the dominant wave signal. .

[0083] At the same time, obtain the coordinates of the sound source. Sensor coordinates and reflective interface Location information. Calculate the beam emitted by the sound source through the reflecting interface. Reflected to the sensor The total length of the geometric polygonal line is used to obtain the theoretical reflection path. Using formulas Calculate the expected arrival time. Then, using... A specific time window is set with the center moment and a preset time delay fluctuation tolerance. This is applied to the main wave signal. Within that specific time window afterwards, the comparison sequence The arrival time of each wave packet component is determined. The wave packet component with the highest degree of overlap with this time window range is identified as the scattered wave signal. .

[0084] Step 404: Compare the start trigger time of the scattered wave signal and the main wave signal under each abnormal path to obtain the time offset corresponding to each abnormal path, and arrange all the time offsets into a time delay sequence according to the channel order in the acoustic emission waveform sequence.

[0085] In this step, the initial trigger time refers to the point in time when the voltage envelope of the wave packet component first exceeds the preset trigger threshold. The time offset is the difference between the initial trigger time of the scattered wave signal and the initial trigger time of the main wave signal. The time delay sequence is a set of values ​​composed of the time offsets corresponding to all abnormal paths arranged in the physical order of the sensor channels.

[0086] In this embodiment, the energy envelope detection algorithm is first used to determine the main wave signal under each abnormal path. The start triggering time and scattered wave signals The start triggering time Next, the time offset corresponding to each abnormal path is calculated using subtraction. Then, all time offsets are arranged according to the physical order of the sensor channels. This ultimately yields a time delay sequence representing the scattering hysteresis characteristics. .

[0087] For example, in a four-channel monitoring scenario, if only channel 1 and channel 3 are abnormal paths, the resulting time delay sequence... , where the value 0 represents a placeholder for a non-abnormal path channel.

[0088] Step 105: Determine the coordinate position of the scattered wave signal based on the time delay sequence, and project the energy mutation point and coordinate position onto the pre-constructed blade model for boundary fitting to determine the physical boundary of the delamination crack in the wind turbine blade.

[0089] In this step, coordinate position refers to the geometric coordinates in space of the scattering source or reflection point obtained by inversion based on the time offset. The pre-built blade model refers to pre-stored digital mesh data reflecting the blade's external profile and internal structure. Delamination crack refers to peeling damage occurring between the composite material layers of the blade. The physical boundary refers to the closed outer contour used to describe the actual extension range and shape characteristics of the delamination crack.

[0090] like Figure 3 As shown, Figure 3A flowchart illustrating the method for determining the physical boundary of delamination cracks in wind turbine blades provided in this application embodiment.

[0091] Step 501: Calculate the product of each time offset in the time delay sequence and the preset sound speed to obtain the travel increment corresponding to each scattered wave signal.

[0092] In this step, the travel increment refers to the increase in the length of the scattered wave propagation path relative to the length of the main wave direct propagation path.

[0093] In this embodiment of the application, the time delay sequence is first extracted. For example, in a scenario with four acquisition channels, the time delay sequence is obtained. Next, obtain the preset sound velocity. The preset sound velocity refers to the theoretical propagation speed constant of elastic waves in the structure, pre-set based on the material properties of the wind turbine blades. Then, the travel increment corresponding to each scattered wave signal is calculated using multiplication operations. The specific calculation formula is: ,in, Indicates the first The travel increment corresponding to each channel Indicates the first Time offset of each channel This indicates the preset speed of sound.

[0094] For example, when the time offset is a numerical value And the preset speed of sound is When the travel increment is obtained, it is the product of the two.

[0095] Step 502: Based on each stroke increment, construct multiple elliptical trajectory lines with the sound source coordinates and the corresponding sensor position as the focus, and take the intersection of multiple elliptical trajectory lines as the coordinate position of each scattered wave signal.

[0096] In this step, the elliptical trajectory refers to a geometric trajectory in three-dimensional space with the sound source position and the sensor position as foci, and the sum of the distances from the moving point to the two foci being a constant. The coordinate position refers to the three-dimensional value of the scattering source in the blade spatial coordinate system, determined by the intersection of multiple elliptical trajectory lines.

[0097] In this embodiment of the application, the coordinates of the sound source are first obtained. and the location of the corresponding sensors Calculate the coordinates of the sound source. With sensor position Geometric distance between Next, with and As the focus, and with An elliptical trajectory is constructed with the major axis as the reference length. Then, a nonlinear equation solving algorithm is used to find the common intersection points of multiple elliptical trajectories. Specifically, the calculation results are interpreted as coordinate positions. .

[0098] For example, coordinate position can be represented as a three-dimensional vector. .

[0099] Step 503: Establish the correlation between each energy mutation point and each coordinate position, and match each coordinate point that meets the correlation to the nearest grid node in the blade model to obtain the projected coordinates corresponding to each grid node, and determine the target connected region based on the projected coordinates.

[0100] In this step, the correlation refers to the spatial distance matching criterion used to determine whether the energy mutation point and the coordinate position belong to the same physical defect. Projected coordinates refer to the grid node position information corresponding to the coordinate position after mapping to the blade's digital model. The target connected region refers to a closed region composed of a set of spatially adjacent grid nodes with damage attributes.

[0101] In this embodiment of the application, the preset spatial radius is first obtained. The preset spatial radius refers to a pre-set geometric distance threshold used to determine the correlation between points. Calculate each energy mutation point. With each coordinate position The Euclidean distance between them. If the geometric spacing is less than... If a correlation is established between the two, then each coordinate point that matches the correlation is matched to the blade model. Specifically, this is done by searching the blade model for coordinates that match the given coordinates. Obtain the projected coordinates from the nearest grid node. Then, based on the projected coordinates The distribution of points is analyzed using region growing to identify adjacent groups of projected points. Finally, the identified set of mesh nodes is determined as the target connected region. .

[0102] Step 504: Perform contour tracing on the outermost edge of the target connected region to obtain closed lines, and use the geometric region enclosed by the closed lines as the physical boundary.

[0103] In this step, a closed line refers to a closed geometric line formed by sequentially connecting boundary nodes determined by contour tracing. A geometric region refers to a planar area enclosed by closed lines in three-dimensional space. A physical boundary refers to the final limit representing the actual extension shape and spatial extent of the layered crack in the blade structure.

[0104] In this embodiment of the application, the target connected region is first identified. All mesh nodes involved are considered, and the outermost edge nodes of the region are extracted. Next, contour tracing is performed on these edge nodes according to a preset path search step size. Then, a spline curve fitting algorithm is used to connect the traced nodes to obtain closed lines. Specifically, the geometric region enclosed by the closed lines represents the actual damaged surface of the delamination crack. Finally, this region is determined as the physical boundary. .

[0105] Step 1051: Calculate the target peak amplitude of all scattered wave signals corresponding to the abnormal path, divide each target peak amplitude into a preset amplitude interval, count the number of scattered waves in each amplitude interval, and obtain the frequency sequence.

[0106] In this step, the target peak amplitude refers to the maximum absolute voltage value reflecting energy intensity extracted from the scattered wave signal corresponding to the abnormal path. The preset amplitude interval refers to a continuous voltage amplitude interval segment pre-divided for performing statistical distribution analysis. The frequency sequence refers to a numerical vector formed by arranging the number of occurrences of the scattered wave signal in each amplitude interval in interval order.

[0107] In this embodiment, firstly, all scattered wave signal waveforms corresponding to the abnormal path are extracted, and the target peak amplitude corresponding to each signal is obtained. For example, the target peak amplitude vector is obtained. Next, the preset amplitude range is retrieved. Next, the number of scattered waves within each amplitude interval was counted. Finally, the number of scattered waves was arranged in ascending order of amplitude value to obtain the frequency sequence. For example, a frequency sequence can be represented as a row vector. .

[0108] Step 1052: Perform linear fitting on the number of scattered waves and the corresponding peak amplitude in the frequency sequence to obtain the slope of the damage distribution.

[0109] In this step, the damage distribution slope refers to the logarithmic decay rate parameter, obtained using a linear fitting algorithm, which describes the frequency variation with amplitude level.

[0110] In this embodiment of the application, the frequency sequence is first... A logarithmic transformation is performed on the center value of the corresponding amplitude interval. The center value of the amplitude interval is used as the independent variable. The natural logarithm of the corresponding frequency is used as the dependent variable. Then, the least squares method is used to perform linear fitting. The specific calculation formula is shown in formula (4): (4) in, Indicates the slope of the damage distribution. Indicates the total number of intervals. Indicates the first The amplitude parameters of each interval This represents the corresponding frequency logarithmic value. The calculated slope of the damage distribution. This indicates the level of activity in the internal damage evolution of the material.

[0111] Step 1053: Compare the damage distribution slope with the preset crack classification threshold to determine the crack damage type corresponding to the delamination crack, and mark the crack damage type to the physical boundary. The crack damage type includes matrix cracking, fiber debonding, and fiber breakage.

[0112] In this step, the preset crack classification threshold refers to the critical index of damage distribution slope used to distinguish different failure modes of composite materials. Crack damage type refers to the specific physical failure category of delamination damage determined based on acoustic statistical characteristics.

[0113] In this embodiment, the calculated damage distribution slope is first retrieved. Next, retrieve the preset crack classification threshold set. Then, set the slope... The comparison is performed with different preset numerical ranges. Specifically, if... If it is located in the first interval, then the crack damage type is determined to be matrix cracking. If If it is located in the second interval, it is determined to be fiber debonding. If If the crack is located in the third interval, it is identified as fiber fracture. Finally, the determined crack damage type is marked at the physical boundary. In the corresponding model attribute fields.

[0114] This application's embodiments ensure high consistency of the original acoustic signal across the time scale through multi-channel synchronous acquisition and analog-to-digital conversion within a preset time window. It resists background noise interference in the blade operating environment and restores the true physical parameters representing the degree of damage. It achieves precise point-to-point localization of the damage initiation location, mitigating the bias in energy spatial distribution assessment in existing technologies. It extracts key time-dependent information carrying damage geometry from complex reflection signals. It achieves a leap from point localization to surface characterization, completing a precise quantitative reconstruction of the physical contour of the damaged area.

[0115] Figure 4 This is a schematic diagram of a specific embodiment of the wind turbine blade crack detection system utilizing acoustic emission signals provided in this application, with reference to... Figure 4 The system may include: The acquisition module 21 is used to acquire acoustic emission signals of wind turbine blades collected by multiple sensors within a preset time window, and to perform synchronous analog-to-digital conversion on each acoustic emission signal to obtain an acoustic emission waveform sequence. The estimation module 22 is used to estimate the time delay of the acoustic emission waveform sequence using cross-correlation to obtain the arrival time difference sequence, and to calculate the distance weight and gain compensation of multiple peak amplitudes extracted from the acoustic emission waveform sequence to obtain the attenuation coefficient matrix. The calculation module 23 is used to calculate the coordinates of the arrival time difference sequence using hyperbolic positioning to obtain the sound source coordinates, and calculate the propagation distance sequence between the sound source coordinates and the sensor. By curve fitting the propagation distance sequence and the attenuation coefficient matrix, the energy distribution matrix is ​​obtained. The decomposition module 24 is used to perform waveform decomposition on the acoustic emission waveform sequence corresponding to the abnormal path in the energy distribution matrix using wavelet transform, to obtain the main wave signal and the scattered wave signal, and to calculate the time offset of the scattered wave signal relative to the main wave signal to obtain the time delay sequence. The abnormal path is determined based on the energy abrupt change point in the energy distribution matrix. The fitting module 25 is used to determine the coordinate position of the scattered wave signal based on the time delay sequence, and to project the energy mutation point and coordinate position onto the pre-built blade model for boundary fitting to determine the physical boundary of the layered crack in the wind turbine blade.

[0116] The wind turbine blade crack detection system utilizing acoustic emission signals in this application embodiment is used to implement the aforementioned wind turbine blade crack detection method utilizing acoustic emission signals. Therefore, the specific implementation of the wind turbine blade crack detection system utilizing acoustic emission signals can be found in the embodiment section of the wind turbine blade crack detection method utilizing acoustic emission signals mentioned above. The specific implementation can be referred to the description of the corresponding embodiments, which will not be repeated here.

[0117] Figure 5 A schematic diagram of the hardware structure of the electronic device provided in an embodiment of this application is shown.

[0118] This application also provides an electronic device, including: a memory for storing a computer program; and a processor for executing the computer program to implement the steps of the wind turbine blade crack detection method using acoustic emission signals described above.

[0119] The electronic device may include a processor 510 and a memory 520 storing computer program instructions.

[0120] Specifically, the processor 510 may include a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits that can be configured to implement the embodiments of this application.

[0121] Memory 520 may include mass storage for data or instructions. For example, and not limitingly, memory 520 may include a hard disk drive (HDD), floppy disk drive, flash memory, optical disk, magneto-optical disk, magnetic tape, or Universal Serial Bus (USB) drive, or a combination of two or more of these. Where appropriate, memory 520 may include removable or non-removable (or fixed) media. Where appropriate, memory 520 may be internal or external to the integrated gateway disaster recovery device. In a particular embodiment, memory 520 is non-volatile solid-state memory.

[0122] Memory may include read-only memory (ROM), random access memory (RAM), disk storage media devices, optical storage media devices, flash memory devices, and electrical, optical, or other physical / tangible memory storage devices. Therefore, typically, memory includes one or more tangible (non-transitory) computer-readable storage media (e.g., memory devices) encoded with software including computer-executable instructions, and when the software is executed (e.g., by one or more processors), it is operable to perform the operations described with reference to the method according to the first aspect of this disclosure.

[0123] The processor 510 reads and executes computer program instructions stored in the memory 520 to implement any wind turbine blade crack detection method utilizing acoustic emission signals in the above embodiments.

[0124] In one example, the electronic device may also include a communication interface 530 and a bus 540. Wherein, such as Figure 5 As shown, the processor 510, memory 520, and communication interface 530 are connected through bus 540 and complete communication with each other.

[0125] The communication interface 530 is mainly used to realize communication between various modules, devices, units and / or equipment in the embodiments of this application.

[0126] Bus 540 includes hardware, software, or both, that couples components of an online data traffic metering device together. For example, and not limitingly, the bus may include an Accelerated Graphics Port (AGP) or other graphics bus, an Enhanced Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), HyperTransport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an Infinite Bandwidth Interconnect, a Low Pin Count (LPC) bus, a memory bus, a Microchannel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local (VLB) bus, or other suitable buses, or combinations of two or more of these. Where appropriate, bus 540 may include one or more buses. Although specific buses are described and illustrated in embodiments of this application, any suitable bus or interconnect is contemplated herein.

[0127] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of any of the above-described methods for detecting cracks in wind turbine blades using acoustic emission signals.

[0128] In one exemplary embodiment, the aforementioned computer-readable storage medium may include, but is not limited to, various media capable of storing computer programs, such as USB flash drives, read-only memory, random access memory, portable hard drives, magnetic disks, or optical disks.

[0129] Embodiments of the present invention also provide a computer program product, which includes a computer program that, when executed by a processor, implements the steps in any of the embodiments of the wind turbine blade crack detection method using acoustic emission signals.

[0130] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0131] The method and system for detecting cracks in wind turbine blades using acoustic emission signals provided in this application have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the embodiments above are only for the purpose of helping to understand the method and its core ideas. It should be noted that those skilled in the art can make several improvements and modifications to this application without departing from the principles of this application, and these improvements and modifications also fall within the protection scope of this application.

Claims

1. A method for detecting cracks in wind turbine blades using acoustic emission signals, characterized in that, include: Acoustic emission signals of wind turbine blades collected by multiple sensors within a preset time window are acquired, and each acoustic emission signal is synchronously converted from analog to digital to obtain an acoustic emission waveform sequence. The time delay of the acoustic emission waveform sequence is estimated by using cross-correlation to obtain the arrival time difference sequence, and the distance weighting and gain compensation are performed on multiple peak amplitudes extracted from the acoustic emission waveform sequence to obtain the attenuation coefficient matrix. The coordinates of the time difference of arrival sequence are calculated using hyperbolic positioning to obtain the coordinates of the sound source, and the propagation distance sequence between the sound source coordinates and the sensor is calculated. The energy distribution matrix is ​​obtained by curve fitting the propagation distance sequence and the attenuation coefficient matrix. Wavelet transform is used to decompose the acoustic emission waveform sequence corresponding to the abnormal path in the energy distribution matrix to obtain the main wave signal and the scattered wave signal. The time offset of the scattered wave signal relative to the main wave signal is calculated to obtain the time delay sequence. The abnormal path is determined according to the energy abrupt change point in the energy distribution matrix. The coordinate position of the scattered wave signal is determined based on the time delay sequence, and the energy mutation point and the coordinate position are associated and projected onto the pre-constructed blade model for boundary fitting to determine the physical boundary of the layered crack in the wind turbine blade.

2. The method according to claim 1, characterized in that, After determining the coordinate position of the scattered wave signal based on the time delay sequence, associating the energy abrupt change point with the coordinate position, and projecting it onto a pre-constructed blade model for boundary fitting to determine the physical boundary of the delamination crack in the wind turbine blade, the method further includes: The target peak amplitude of all scattered wave signals corresponding to the abnormal path is counted, and each target peak amplitude is divided into a preset amplitude interval. The number of scattered waves in each amplitude interval is counted to obtain a frequency sequence. The slope of the damage distribution is obtained by linearly fitting the number of scattered waves and the corresponding peak amplitude in the frequency sequence. The slope of the damage distribution is compared with a preset crack classification threshold to determine the crack damage type corresponding to the delamination crack, and the crack damage type is marked to the physical boundary. The crack damage type includes matrix cracking, fiber debonding, and fiber breakage.

3. The method according to claim 1, characterized in that, The time delay of the acoustic emission waveform sequence is estimated using cross-correlation to obtain the arrival time difference sequence, including: The channel with the earliest arrival time in the acoustic emission waveform sequence is taken as the reference channel, the waveform corresponding to the reference channel is taken as the reference waveform, and the remaining channels in the acoustic emission waveform sequence are taken as the channels to be tested, and the channels corresponding to the channels to be tested are taken as the waveforms to be tested. Each waveform to be measured is translated, and the degree of overlap between the envelope of each waveform to be measured and the reference waveform is compared to obtain the relative displacement of each waveform to be measured. Each relative displacement is multiplied by a preset sampling period to obtain the relative time delay value of each waveform under test. All relative time delay values ​​are then arranged according to the channel order in the acoustic emission waveform sequence to obtain the arrival time difference sequence.

4. The method according to claim 3, characterized in that, Distance weighting and gain compensation are performed on multiple peak amplitudes extracted from the acoustic emission waveform sequence to obtain an attenuation coefficient matrix, including: The highest voltage point of each channel in the acoustic emission waveform sequence is taken as the peak amplitude of the waveform of each channel; The relative time delay value in the arrival time difference sequence is multiplied by the preset wave velocity value to obtain the distance deviation value of each channel under test relative to the reference channel. Based on each distance deviation value and the spatial position of the corresponding sensor, multiple sets of positioning trajectory lines are determined, and the intersection point of the multiple sets of positioning trajectory lines is determined as the initial sound source of the acoustic emission signal. Calculate the geometric distance between the initial sound source and each sensor, and based on the geometric distance, correct each peak amplitude to obtain the corrected peak amplitude; The quotient between the corrected peak amplitude corresponding to the reference channel and the corrected peak amplitude corresponding to each channel under test is calculated to obtain the attenuation coefficient matrix.

5. The method according to claim 4, characterized in that, Hyperbolic positioning is used to calculate the coordinates of the time difference of arrival sequence to obtain the sound source coordinates. The propagation distance sequence between the sound source coordinates and the sensor is then calculated. By curve fitting the propagation distance sequence and the attenuation coefficient matrix, an energy distribution matrix is ​​obtained, including: Based on the arrival time difference sequence, multiple sets of hyperbolic trajectory equations are constructed with any two sensors as the focus, and multiple candidate points are selected within the area determined by the multiple sets of hyperbolic trajectory equations with the initial sound source as the search center. Calculate the theoretical time delay difference between each candidate point and the sensor corresponding to the channel under test and the sensor corresponding to the reference channel. Calculate the difference between each theoretical time delay difference and the relative time delay value of the corresponding channel in the arrival time difference sequence to obtain the residual corresponding to each candidate point. Then, take one or more candidate points corresponding to the minimum residual as the intersection point. Calculate the spatial centroid of all intersection points to obtain the sound source coordinates, calculate the straight line length between the sound source coordinates and each sensor to obtain the propagation distance sequence, and construct a fitting function with each straight line length in the propagation distance sequence as the independent variable and the quotient of the corresponding channel in the attenuation coefficient matrix as the dependent variable. Using the spatial location of each grid node in the pre-constructed blade model as a sampling point, the coordinates corresponding to each sampling point are substituted into the fitting function for assignment and mapping to obtain the energy distribution matrix.

6. The method according to claim 5, characterized in that, Wavelet transform is used to decompose the acoustic emission waveform sequence corresponding to the abnormal path in the energy distribution matrix to obtain the main wave signal and the scattered wave signal. The time offset of the scattered wave signal relative to the main wave signal is calculated to obtain the time delay sequence, including: The numerical differences between adjacent coordinate points in the energy distribution matrix are calculated to obtain multiple numerical change rates. The coordinate points corresponding to the numerical change rates that exceed the preset deviation are determined as energy mutation points. Based on the structural discontinuity of the region where each energy mutation point is located, the reflection interface corresponding to each energy mutation point is determined. The line connecting the sound source coordinates to each sensor is used as a candidate path, and the candidate path that intersects with any energy mutation point is determined as an abnormal path. Frequency band layering is performed on the acoustic emission waveform sequence corresponding to each abnormal path to obtain multiple wave packet components corresponding to each abnormal path. The earliest arriving wave packet component in each abnormal path is taken as the main wave signal, and the wave packet component corresponding to the reflection interface is taken as the scattered wave signal. The starting trigger time of the scattered wave signal under each abnormal path is compared with that of the main wave signal to obtain the time offset corresponding to each abnormal path. Then, all time offsets are arranged into a time delay sequence according to the channel order in the acoustic emission waveform sequence.

7. The method according to claim 1, characterized in that, The coordinate position of the scattered wave signal is determined based on the time delay sequence, and the energy abrupt change point and the coordinate position are correlated and projected onto a pre-constructed blade model for boundary fitting to determine the physical boundary of the delamination crack in the wind turbine blade, including: Calculate the product of each time offset in the time delay sequence and the preset sound speed to obtain the travel increment corresponding to each scattered wave signal; Based on each stroke increment, multiple elliptical trajectory lines are constructed with the sound source coordinates and the corresponding sensor position as the focus, and the intersection of multiple elliptical trajectory lines is used as the coordinate position of each scattered wave signal. Establish the correlation between each energy mutation point and each coordinate position, and match each coordinate point that conforms to the correlation to the nearest grid node in the blade model to obtain the projected coordinates of each grid node. Determine the target connected region based on the projected coordinates. The outermost edge of the target connected region is traced to obtain a closed line, and the geometric region enclosed by the closed line is used as the physical boundary.

8. A wind turbine blade crack detection system utilizing acoustic emission signals, characterized in that, include: The acquisition module is used to acquire acoustic emission signals of wind turbine blades collected by multiple sensors within a preset time window, and to perform synchronous analog-to-digital conversion on each acoustic emission signal to obtain an acoustic emission waveform sequence. The estimation module is used to estimate the time delay of the acoustic emission waveform sequence using cross-correlation to obtain the arrival time difference sequence, and to calculate the distance weight and gain compensation of multiple peak amplitudes extracted from the acoustic emission waveform sequence to obtain the attenuation coefficient matrix. The calculation module is used to calculate the coordinates of the arrival time difference sequence using hyperbolic positioning to obtain the sound source coordinates, and calculate the propagation distance sequence between the sound source coordinates and the sensor. By performing curve fitting on the propagation distance sequence and the attenuation coefficient matrix, the energy distribution matrix is ​​obtained. The decomposition module is used to perform waveform decomposition on the acoustic emission waveform sequence corresponding to the abnormal path in the energy distribution matrix using wavelet transform to obtain the main wave signal and the scattered wave signal, and to calculate the time offset of the scattered wave signal relative to the main wave signal to obtain the time delay sequence. The abnormal path is determined according to the energy abrupt change point in the energy distribution matrix. The fitting module is used to determine the coordinate position of the scattered wave signal based on the time delay sequence, and to project the energy mutation point and the coordinate position onto the pre-constructed blade model for boundary fitting to determine the physical boundary of the delamination crack in the wind turbine blade.

9. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor, configured to execute the computer program to implement the steps of the wind turbine blade crack detection method using acoustic emission signals as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, enables the wind turbine blade crack detection method using acoustic emission signals as described in any one of claims 1 to 7.

Images